1. Field of the Invention
The present invention relates to a substrate processing apparatus checking method and a storage medium, and more particularly to a method of checking a substrate processing apparatus using time-series data on parameters relating to processing carried out on substrates by the substrate processing apparatus.
2. Description of the Related Art
A substrate processing apparatus such as an etching apparatus or a CVD apparatus is manufactured by an apparatus manufacturing company (manufacturer), then disassembled once and delivered to an apparatus using company (customer), and further assembled by the apparatus using company. In general, it is necessary for the apparatus manufacturing company to assure the quality of the substrate processing apparatus, and for quality assurance, it is necessary to prove that the performance of the delivered and assembled substrate processing apparatus is the same as or rarely different from the as-manufactured performance of the substrate processing apparatus. In the case that a plurality of substrate processing apparatuses of the same model are delivered, it is necessary for the apparatus manufacturing company to prove that there is no little difference in performance between the substrate processing apparatuses.
To prove such things, a substrate processing apparatus delivered and assembled is operated once to carry out processing on substrates, time-series data on parameters relating to the processing is sampled, and the substrate processing apparatus is checked using the sampled data.
As such a checking method, there is known a method in which the emission intensity of plasma in plasma processing on wafers is sampled as operation data, and multivariable analysis (mainly, principal component analysis) is carried out using the operation data to detect abnormalities in the apparatus (see e.g. Japanese Laid-Open Patent Publication (Kokai) No. 2004-47501).
In recent years, with miniaturization of semiconductor devices manufactured from wafers, the transient response characteristics of the apparatus (for example, the time from when the application of radio-frequency voltage from a lower electrode of the apparatus is started until when the applied radio-frequency voltage becomes stable) is required to be additionally managed, and hence the transient response characteristics of the apparatus as well are required to be checked.
On the other hand, in the above described checking method, because multivariable analysis is carried out, the effects of operation data throughout the whole period of plasma processing are reflected on obtained checking results, and transient data changes cannot be checked.
To cope with this, there is known a method in which time-series data on parameters relating to plasma processing throughout the whole period of plasma processing as a reference is set as a set value pattern, and in plasma processing carried out by a substrate processing apparatus, time-series data on parameters relating to the plasma processing is sampled and set as a measured value pattern, and the set value pattern and the measured value pattern are directly compared with each other (see e.g. Japanese Laid-Open Patent Publication (Kokai) No. 2000-269108). In this method, because statistical processing such as multivariable analysis is not performed on data to be compared, transient data changes (transient response characteristics) can be checked.
However, the set value pattern described above is time-series data throughout the whole period of the plasma processing, setting the set value pattern takes much time and effort, and comparison between the set value pattern and the measured value pattern puts a heavy load on a PC or the like.
From the viewpoint of reducing man-hour for comparison, part of the set value pattern and part of the measured value pattern are sometimes compared with each other, but extraction (narrowing), comparison, etc. of data from the set value pattern and so on are manually carried out by an operator, which causes human errors and lowers the reliability of checking of transient response characteristics. Further, there are many situations where a simple analyzing process is manually carried out on data extracted at the time of comparison, but in such situations, it is necessary to define the contents of the analyzing process (define macros and calculation formulas using spreadsheet software) as occasion demands, and as a result, it takes much time to check transient response characteristics.